A stereoscopic image refers to a 3-dimensional (3D) image that provides image data together with depth data and shape data regarding a space. While a stereo image provides images respectively corresponding to difference viewpoints of the left eye and the right eye, a stereoscopic image provides an image as if the image is viewed from a different location when a viewer changes his/her viewpoint. Therefore, images captured at various viewpoints are required to generate a stereoscopic image.
Images captured at various viewpoints to generate a stereoscopic image constitute a massive amount of data. Therefore, in consideration of network infrastructures and the ground wave bandwidth, it is almost impossible to display a stereoscopic image even if the stereoscopic image is coded by using an encoding apparatus optimized to single-view video encoding techniques, such as MPEG-2, H.265/AVC, and HEVC.
Therefore, a multiview (multilayer) image encoding apparatus is required to generate a stereoscopic image. In particular, it is necessary to develop a technique for efficiently reducing redundancy between time points and viewpoints.
For example, a multiview video codec may improve compression efficiency by compressing a basic viewpoint by using a single-view video encoding technique and encoding expanded viewpoints with reference to the basic viewpoint. Furthermore, by further encoding additional data like a depth image, an image including more viewpoints than viewpoints input via a decoder may be generated. Here, the depth image is used for synthesizing intermediate viewpoint images instead of being directly displayed to a user, where degradation of the depth image deteriorates the quality of a synthesized image. Therefore, it is necessary for a multiview video codec to efficiently compress not only a multiview video, but also a depth image.